ath11k: 802.11s peer stuck at 1 MPDU per PPDU until DELBA (IPQ8074, WLAN.HK.2.9.0.1-02146)

Andreas Ufert ath11k at domov.de
Thu Nov 12 22:20:25 PST 2026


Hello,

a short follow-up with one more occurrence, this time on TID 7, and
the first one caught by the tid_flags check from my first mail.

After that mail I added a per-minute check that reads tid_flags
(HTT_TX_TID_STATS_V1_TLV) for every peer and TID 0-7 and alerts
when bits 0x4 and 0x8 go from set to clear for three minutes.

13 Nov, 00:30 CET: the check fired for TID 7 on A -> R and B -> R,
in the same minute. TIDs 0 and 1 stayed set on all links, and an
hourly iperf3 probe (TID 0) was healthy all night, so without the
check this would have gone unnoticed. At the onset there was no
"tx failed" burst, nothing in the logs or dmesg, no re-peering, no
reboot. Mesh traffic per node around 00:30: A sent 15-54 Mbit/s
(destination and TID not recorded), B less than 1.2 Mbit/s, R 2-8.
So, as on 10 Nov, a sender that was almost idle went into the state
in the same minute as the other one.

Measured at 06:41, 8 s of iperf3 each, htt_stats type 1 on the
sender, TID 7 via --tos 0xE0:

  sender->peer  TID  Mbit/s   MPDUs  PPDUs  MPDU/PPDU  MSDU/MPDU
  A->R          7      29.7   21728  21828       0.99       1.03
  A->R          0       550  131434   8022      16.4        2.91
  B->R          7      23.2   14646  15063       0.97       1.12
  B->R          0       407  142674   5705      25.0        1.97

tx_xretry over the same runs: A->R TID 7 2311, TID 0 495; B->R
TID 7 5199, TID 0 1053. On R (type 2, during the TID 7 runs): from A
22011 PPDUs, 22300 MPDUs FCS ok, 83 FCS errors; from B 17921, 18015,
243. Again a clean air link without aggregation.

tid_flags for TID 7 to R read 0x1014000 on both senders, all other
TIDs in use 0x101400c. On B it read 0x1014800 once at idle before
the load runs; I have not seen bit 0x800 before. The load did not
change the flags. agg_status on R showed TID 7 as "RX 1, DTKN 0x01,
SSN 0x000" for both peers, the same as on healthy TIDs.

Cure at 06:42:51, on R, both peers at once (no control this time):

  echo "rx stop 7" > .../stations/<A>/agg_status
  echo "rx stop 7" > .../stations/<B>/agg_status

Afterwards TID 7: A -> R 683 Mbit/s, B -> R 372 Mbit/s, tid_flags
0x101400c on both. Right after the DELBA and before the next TID 7
traffic the check still saw the bits clear on B for one more
minute, which fits the reading "aggregation set up" rather than
"healthy".

So far:
- the state has hit TIDs 0, 1 and 7, always per peer and TID;
- twice now (10 Nov and 13 Nov) two senders towards R went into it
  in the same minute, one of them almost idle;
- each time a DELBA for that TID from the receiver cleared it (three
  occurrences, eight peer/TID pairs);
- it lasted about six hours this time, until I intervened.

Full dumps from before and after the cure are available on request.

Thanks,
Andreas Ufert



More information about the ath11k mailing list